You have seen them: a sound, a format or a piece of choreography that was everywhere on TikTok, then on Reels, and eventually, in a suit-and-tie version, on LinkedIn. Some trends seem to never end, and people keep jumping on them long after the first wave.
For most marketing teams, this raises the obvious question: is it still worth joining? And the maybe not so obvious one: how closely should we mimic the trend?
The usual answer is "fast and faithful": move fast and join when the trend is fresh, and reproduce what worked for others. However, new research suggests that this answer might be wrong.
Bravin, Clegg, Hofstetter, Pouly and Berger (2026), publishing in the Journal of Marketing, examined how closely creators should follow a trend and found that videos which deviate from the trend generate more engagement than videos which copy it closely. The effect is causal, it holds for the same person posting both versions, and it gets stronger the later you arrive.
The study caught my attention for a few reasons. I know one of the authors personally, and it is a paper carried largely by Swiss institutions: four of the five authors work in Lucerne, Lausanne and St. Gallen. The fifth is Jonah Berger of Wharton, the author of Contagious and the STEPPS framework I use in my teaching to explain why some content gets shared more. And, last but not least, it sheds light on a question clients have asked me: should we follow this trend, and how?
How the study was built: 86,000 videos and one field experiment
Bravin and colleagues first analysed 86,244 TikTok dance videos across 1,548 trends, measuring how far each video deviated from the others in the same trend and comparing engagement while holding the creator, the trend and the time of posting constant. They then ran two online experiments in which only the choreography of one video was changed. Finally, they recruited 38 real TikTok creators, each of whom posted a typical and an atypical version of the same trend three days apart.
Engagement was measured as a rate (likes per view) rather than as raw likes. This matters because it captures how appealing content is once someone has seen it, independently of how many people the algorithm chose to show it to.
Across all four studies, the answer was the same. More atypical videos receive more engagement.
The core idea: the twist only works if people know the trend
In the observational data, a one standard deviation increase in atypicality was associated with 2.5% higher odds that a view turns into a like. In plain terms, a standard deviation is the typical distance from the average, so this compares an average video with one that is noticeably, but not extremely, more different from its trend. For a median creator with roughly 15,000 views per video, the authors estimate about 69 additional organic likes per post, or around 10,800 over a year of posting three times a week.
In the experiments, the effect is even larger. Participants who watched six videos from the same trend were nearly twice as likely to engage with the one that deviated (24% vs. 13%). In the field experiment, the atypical version posted by the same creator received roughly 10% higher odds of a like.
The interesting part is why. The effect disappears entirely when the audience has not seen the trend before. With high trend exposure, the atypical version was chosen 34% of the time against 18% for the typical one. With low exposure, the two versions performed identically.
Think of a well-known joke told with a different punchline. The laugh depends on the audience knowing the original. Tell it to someone who has never heard the joke, and the new punchline is just a punchline. The authors ground this in what psychologists call expectancy violation: people need to recognise the pattern before they can enjoy the deviation from it.
This is consistent with what we know from other domains. Berger and Packard (2018) found that songs whose lyrics differ more from their genre become more popular. Lysyakov and colleagues (2025), also in the Journal of Marketing, found that retailers whose social media content is less similar to their close competitors' get higher engagement and gain followers faster. What Bravin and colleagues add is the timing, and the timing is where the practical value lies for smaller organisations.
Why this matters for SMBs
Trend-following is usually treated as a speed game that only marketing teams with dedicated social media managers can win. There are three details in the study which the authors mention but do not dwell on, and which, in my opinion, turn trend-following into a timing game instead.
First, late is not lost. The more videos already exist in a trend, the stronger the atypicality effect becomes. Smaller organisations are often late to trends, because content has to pass through people who have other jobs. According to this study, that lateness is only a handicap if you copy the trend faithfully.
Second, the twist belongs at the end. Deviations in the final seconds of a video were associated with higher engagement; deviations in the opening seconds were not. The audience needs the recognisable start to place the trend.
Third, "moderate" is enough. A follow-up experiment found the effect holds even when deviations are only moderate, and the field experiment was designed exactly that way: creators changed one element and kept everything else constant.
One more point, and this one is my reading rather than the authors'. The study measures engagement, not reach. But in today's interest-based social feeds, engagement rate is one of the signals platforms read as "people like this", so a higher engagement rate is likely to translate into more distribution. The study does not test this. It is, however, the mechanism by which engagement becomes visibility.
If you want to understand why organic engagement now decides what the platforms amplify, and why paid media should come last rather than first, I unpacked the shift from the social graph to the interest graph in Why Your Best Content Should Earn Reach Before You Pay for It.
What the study cannot tell you
The evidence is based on dance trends on TikTok, produced by individual creators, with likes as the only engagement measure. Brands appear in the paper's introduction, not in its data, and the authors list other trend types as future research. Applying this to a Swiss SME posting Reels or LinkedIn formats is therefore an extrapolation, resting on a mechanism that has been replicated across songs, retailers and product design rather than on the TikTok numbers themselves.
Two studies help draw the boundaries of that mechanism. The first confirms it from the product side. Landwehr, Wentzel and Herrmann (2013) found that typical car designs are preferred at first sight, while atypical designs only gain appeal with repeated exposure. That is the trend-exposure finding in another setting: novelty pays once people know the norm, and until then the faithful version wins. The second sets a limit the study does not test. Sewak, Lee and Haderlie (2025) found that memes raise engagement for brand posts in general but lower it in cause-related campaigns. The deviation is measured against the trend, but the format still has to fit your message. Together, they tell you when the twist works (once the pattern is known) and when it backfires (when it clashes with what you stand for).
How to follow a trend: Recognise, Replicate, Reinterpret
What follows is my attempt to turn the findings into a sequence a small team can actually follow. It is an organising principle rather than a proven methodology, but each step carries one finding from the study, and together they replace "fast and faithful" with something you can put in a content brief.
- Recognise. Choose a trend your audience already recognises, not the one that appeared yesterday. The question is not "is this trend still hot?" but "has our audience seen this enough times to know what we are playing with?" Being late is not the problem, it is the condition for the twist to be seen. Don't be late like everybody else. Be late with a reason.
- Replicate. Reproduce the recognisable elements faithfully: the sound, the format, the opening, the setup. This is the part most teams are tempted to "make their own" from the first second, and it is precisely the part that should stay untouched.
- Reinterpret. Change one element, and change it towards the end: the ending, the cast, the setting, the moment your product or your people appear. Again, a moderate change is enough. The deviation is measured against the trend, not against your brand. In the study, atypicality meant a change in the content itself (the choreography, the setting, the ending), not a logo or a product placed on top of a faithful copy. Your twist should be a different way of doing the trend that happens to be recognisably yours, not a brand element added to it.
Three things to do when the next trend catches on
Recognise, Replicate, Reinterpret is the mental model. Here is what it looks like in your week: the first two points are the three steps as they show up in the content meeting and the storyboard, the third is how you know it worked.
- Wait, then twist. When a trend comes up in your content meeting, ask whether your customers have already seen it. If they have not, copy it faithfully or wait. If they have, join with a reinterpreted ending. Being late and identical is the weakest position of all.
- Keep the core, change the close. Storyboard the post in two parts: everything up to the last few seconds stays true to the trend, and one element changes at the end. Hold everything else constant, the way the creators in the field experiment did.
- Compare reactions per view, not reactions. When you assess the post, use engagement rate rather than total engagement. That is the comparison the study makes, and the only one that tells you whether the twist worked rather than whether the algorithm was generous that day. If my reading above is right, the two should then move together: a post that earns a higher reaction rate should also be shown to more people. A jump in reach is a welcome sign, but the rate is what tells you the twist did the work.
One last thing: a link to distinctiveness
My reading of this study connects it to a distinction the Ehrenberg-Bass Institute has made popular: differentiation, being different from competitors, versus distinctiveness, being recognisable as yourself. Byron Sharp's work suggests that distinctiveness, not differentiation, is what builds mental availability and, eventually, sales.
Reinterpreting a trend is, in practice, a distinctiveness move. The trend supplies the familiar frame, and your reinterpretation (the ending, the setting, the people) is where your own distinctive assets appear. The twist that earns the engagement is also the moment the content becomes recognisably yours. The study hints at this: its follow-up experiments suggest the effect runs through the video being perceived as more unique and its creator as more creative, which is partly a perception of the sender, not only of the clip.
That is my interpretation, not the authors', and the study was not designed to test it. But if it holds, following a trend the right way does more than earn a few extra likes. It puts your brand's recognisable elements in front of an audience that is paying attention.
You do not need to be first, and you do not need to be loud. You need to know when your audience is ready, keep the part they recognise, and change the part they will remember.
If you find this kind of evidence-based content useful, subscribe to my newsletter, where I break down recent marketing research and translate it into frameworks you can actually use.
I am curious how this plays out in your organisation: have you followed a trend recently, and would this change how you approach the next one?
